Many electronic products need various amounts of memory to store information, e.g. data. One common type of high speed, low cost memory includes dynamic random access memory (DRAM) comprised of individual DRAM cells arranged in arrays. DRAM cells include an access transistor, e.g a metal oxide semiconducting field effect transistor (MOSFET), coupled to a capacitor cell. Another type of high speed, low cost memory includes floating gate memory cells. A conventional horizontal floating gate transistor structure includes a source region and a drain region separated by a channel region in a horizontal substrate. A floating gate is separated by a thin tunnel gate oxide. The structure is programmed by storing a charge on the floating gate. A control gate is separated from the floating gate by an intergate dielectric. A charge stored on the floating gate effects the conductivity of the cell when a read voltage potential is applied to the control gate. The state of cell can thus be determined by sensing a change in the device conductivity between the programmed and un-programmed states.
Dynamic random access memories (DRAMs) are the most cost-effective, high-speed memory used with present day computers. DRAMS last nearly indefinitely. Moreover, DRAMs are available in very high density configurations, e.g., 64 megabytes (MB). However, DRAMS can only store information for a limited time without constant refreshing and lose all knowledge of their state once power is removed.
Ferroelectric capacitors are currently being used as nonvolatile memory devices, as disclosed in U.S. Pat. No. 4,888,733, which is entitled “Nonvolatile Memory Cell and Sensing Method” and discloses a two transistor, one capacitor (2T/1C) memory cell. In addition, ferroelectric capacitors are often employed in nonvolatile random access memories (NVRAMs). Memory cells having structures approximating that of DRAMs, i.e., arranged in the conventional one transistor, one capacitor (1T/1C) memory cell pattern, are disclosed in U.S. Pat. No. 5,600,587 (Ferroelectric Random Access Memory), U.S. Pat. No. 5,572,459 (Voltage Reference for a Ferroelectric 1T/1C Based Memory), U.S. Pat. No. 5,550,770 (Semiconductor Memory Device Having FE Capacitor Memory Cells with Reading, Writing, and Forced Refreshing Functions and a Method of Operating the Same), U.S. Pat. No. 5,530,668 (Ferroelectric Memory Sensing Scheme Using Bit Lines Precharged to a Logic One Voltage), and U.S. Pat. No. 5,541,872 (Folded Bit Line Ferroelectric Memory Device). It should be mentioned that all of the patents cited above are incorporated by reference herein for all purposes.
In the above-mentioned patents, the capacitor in a 1T/1C DRAM structure is replaced by a ferroelectric capacitor, as illustrated in FIG. 1A. The memory cell 1 of FIG. 1A includes a wordline 10, a digitline 12, a transistor 14 and a ferroelectric capacitor 16. In operation, the remnant charges are detected on a bit line in a manner similar to the operation of a conventional DRAM. However, these structures are not without attendant problems.
There is an extensive body of literature on both the electrical characteristics of ferroelectric capacitors and applications in cells similar to conventional DRAMs, except that these cells are classified as nonvolatile RAMs (NVRAMs) instead of DRAMs. One recent publication by K. Asaril et al., entitled “Multi-Level Technologies for FRAM Embedded Reconfigurable Hardware” (IEEE Int. Solid-State Circuits Conf., San Francisco 1999, pp. 108-109), describes the use of a ferroelectric capacitor in a ferroelectric RAM (FRAM)which is employed with low voltage to store and access RAM data superimposed on less-frequently accessed read-only memory (ROM) data in the same cell. See U.S. Pat. No. 5,539,279, entitled “Ferroelectric Memory.” The RAM data is volatile and needs to be refreshed on a regular basis, or this aspect of the cell acts like a DRAM. The electrically alterable read-only memory (EAROM), e.g., an electrically erasable programmable ROM (EEPROM) data is nonvolatile. In other words, the FRAM acts like a DRAM with a “repressed” nonvolatile read only “memory,” or “repressed memory.”
Other applications use the ferroelectric capacitors as part of a stacked gate structure similar not to DRAMs but rather to EEPROM or flash memory devices. See, for example, U.S. Pat. No. 5,541,871 (“Nonvolatile Ferroelectric Semiconductor Memory”) and U.S. Pat. No. 5,856,688 (“Integrated Circuit Memory Devices Having Nonvolatile Single Transistor Unit Cells Therein”), which patents are also incorporated herein by reference for all purposes. It will be appreciated that in these devices, however, the remnant charge or polarization charge of the ferroelectric element is used to store information rather than electrons injected onto or removed from a floating gate by hot electron effects and/or tunneling. Instead, the charge differences are not differences in the number of electrons trapped on the gate but rather the polarization charge of the upper ferroelectric capacitor.
It should also be mentioned that the prior art referenced in U.S. Pat. No. 5,541,871 describes a basic structure consisting of a ferroelectric capacitor in series with a gate capacitance, which are both planar structures. As shown in FIG. 1B, a memory cell 1′ includes a wordline 10, a digitline 12, a transistor 14, and a ferroelectric capacitor 20, one plate of the ferroelectric capacitor forming the gate of the transistor 14.
FIG. 1C illustrates a memory cell 1″, which includes the wordline 10, digitline 12, transistor 14, a high dielectric capacitor 30 and a ferroelectric capacitor 32. It will be noted that the lower plate of capacitor 30 is the channel of the transistor 14.
The problem with the above disclosed structure is that the ferroelectric elements have very high electric relative permittivities as, for instance, 80 and 150. It will be appreciated that these permittivity values are 20 to 40 times higher than that of silicon dioxide. It will also be appreciated that, if comparable thicknesses of materials as are used in constructing the ferroelectric and conventional capacitors, only a small fraction, e.g., 2% to 5%, of the voltage applied across the series capacitors will appear across the ferroelectric capacitor. Thus, if the ferroelectric capacitor has a coercive voltage, Vc, i.e., the voltage required for programming, of 1 V or 3 V, then the total word line voltage required for programming the memory cell will be on the order 20 V to 150 V. It will be noted that these are far in excess of voltages used on current CMOS-integrated circuits.
U.S. Pat. No. 5,856,688 seeks to solve this problem by using a “C” shaped floating gate and two control gates. One control gate is a plate of the ferroelectric capacitor and the other control gate is a plate of another capacitor fabricated using a high dielectric constant material. The two capacitors in series, one with a high dielectric constant ferroelectric and the other with a high dielectric constant insulator, are used to program the ferroelectric capacitor at lower voltages. Since the capacitances are more or less comparable, the programming voltage will divide more equally, resulting in a significant fraction across the ferroelectric capacitor.
It should be mentioned that other repressed memory devices have been proposed. For example, a repressed memory where the NVRAM function is provided by a flash memory type structure for the gate of the transfer device is described in the commonly assigned, copending application entitled “DRAM AND SRAM MEMORY CELLS WITH REPRESSED MEMORY” (Ser. No. (09/362,909, filed Jul. 29, 1999), which application is incorporated herein by reference for all purposes. It will be appreciated that these memories do not function like the shadow RAM disclosed in U.S. Pat. 5,399,516 (“Method of Making a Shadow Ram Cell Having a Shallow Trench EEPROM”), storing the same information on both memory planes.
In contrast, U.S. application Ser. No. 09/385,380, describes a memory cell 1′″ having first and second operating modes. The memory cell 1′″ includes a charge transfer transistor 14 having a gate adjacent to a channel region coupling source and drain regions, a digitline 12 coupled to one of the source and drain regions, a storage capacitor 16 coupled to the other of the source and drain regions, a ferroelectric capacitor 32, and a wordline 10 coupled to the gate by the ferroelectric capacitor 32. The polysilicon vertical interconnect connecting the ferroelectric capacitor 32 with oxide layer covering the gate of the transistor 14 forms an intrinsic capacitor 30. (See FIG. 1D). Data is written to and read out of the storage capacitor during the first operating mode, i.e., a normal DRAM access speeds, and written to and read out of the ferroelectric capacitor during the second mode of operation, i.e., at speeds several times slower than normal DRAM access speeds.
It would be very desirable if a memory cell or device could be developed with all of the positive features of DRAMS, i.e., cost, size, speed, availability, etc., which also is nonvolatile, i.e., maintains its memory state with power removed. It would be highly desirable to have a memory element and corresponding system that functions at speeds comparable to that of today's DRAMs. It would also be desirable if at least one logic device could be implemented using a modified form of the memory cell.
With successive generations of DRAM chips, an emphasis continues to be placed on increasing array density and maximizing chip real estate while minimizing the cost of manufacture. It is further desirable to increase array density with little or no modification of the DRAM optimized process flow.
A requirement exists for memory devices which need only be programmed once, as for instance to function as an electronic film in a camera. If the memory arrays have a very high density then they can store a large number of very high resolution images in a digital camera. If the memory is inexpensive then it can for instance replace the light sensitive films which are used to store images in conventional cameras.
Thus, there is a need for improved DRAM technology compatible write once read only memory. It is desirable that such write once read only memory be fabricated on a DRAM chip with little or no modification of the DRAM process flow. It is further desirable that such write once read only memory operate with lower programming voltages than that used by conventional DRAM cells, yet still hold sufficient charge to withstand the effects of parasitic capacitances and noise due to circuit operation.